Inflation Prediction: A Hybrid Time-Series Approach

Abstract

Accurately forecasting inflation is a vital aspect of economic strategy. However, it presents challenges due to its complex and often nonlinear nature, which is influenced by a range of external factors. This study explores an integrated modelling framework that leverages both traditional time series analysis and modern techniques to improve inflation prediction. Using economic data from Ireland and the United Kingdom, four hybrid models were developed by combining Seasonal Autoregressive Integrated Moving Average with Exogenous Variables (SARIMAX) with machine learning and deep learning algorithms, namely, Random Forest, Support Vector Regression, and Long Short-Term Memory networks. Among these, the Seasonal Autoregressive Integrated Moving Average with Exogenous Variables and Long-Short Term Memory (SARI-LSTM) model delivered the most consistent performance across key evaluation metrics, Mean Absolute Estimate (MAE), Root Mean Square Estimate (RMSE), and Mean Absolute Percentage Estimate (MAPE), effectively capturing both seasonal trends and sequential patterns in the data. The results highlight the benefit of combining traditional statistical techniques with modern modelling approaches to produce more reliable and interpretable forecasts. This method offers policymakers and economists valuable insights for managing the uncertainties of inflation.